Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A computer-implemented method, comprising: receiving a compositional constraint from a user for merging multiple images to create a canvas; finding a vector for the canvas in a merged space associated with the compositional constraint; generating a synthetic image for the canvas based on the vector for the canvas and a generative tool trained in a generative adversarial configuration against a discriminative tool; and providing the synthetic image for the canvas to a user when the discriminative tool identifies the synthetic image as a real image, wherein generating a synthetic image for the canvas comprises merging a first saliency measure for a first image in the canvas with a second saliency measure for a second image in the canvas to form a compound saliency measure associated with the vector for the canvas in the merged space.
2. The computer-implemented method of claim 1 , wherein receiving a compositional constraint comprises receiving a positional constraint within the canvas for at least one of the images.
This invention relates to computer-implemented methods for generating image compositions, particularly addressing the challenge of arranging multiple images within a defined canvas while adhering to specific positional constraints. The method involves receiving a compositional constraint, which includes a positional constraint that specifies the desired location of at least one image within the canvas. The system then generates a composition by arranging the images according to these constraints, ensuring that the specified image is placed in the designated position while optimizing the overall layout. The method may also incorporate additional constraints, such as size, orientation, or spatial relationships between images, to produce a visually balanced and aesthetically pleasing arrangement. The invention is particularly useful in applications like photo collage creation, graphic design, and automated layout generation, where precise control over image placement is required. The system dynamically adjusts the remaining images to fit within the canvas while respecting the positional constraints, ensuring a coherent and well-structured composition. This approach enhances user control and automation in image arrangement tasks.
3. The computer-implemented method of claim 1 , wherein generating a synthetic image for the canvas comprises relaxing a positional condition for at least two images within the canvas.
This invention relates to computer-implemented methods for generating synthetic images, particularly in the context of image composition or collage creation. The problem addressed is the challenge of seamlessly integrating multiple images into a single canvas while maintaining visual coherence and avoiding positional conflicts that may disrupt the final composition. The method involves generating a synthetic image by relaxing positional constraints for at least two images within the canvas. This means that the system dynamically adjusts the placement or alignment of these images to resolve conflicts, such as overlapping regions or misalignment, ensuring a more natural and aesthetically pleasing result. The relaxation process may involve techniques like spatial optimization, blending, or transformation to harmonize the images without strict adherence to their original positions. The broader method includes capturing or selecting multiple images, analyzing their spatial relationships, and applying transformations to integrate them into a unified canvas. The relaxation of positional conditions is a key step in this process, allowing the system to dynamically adjust image placements to avoid conflicts while preserving the overall composition's integrity. This approach is particularly useful in applications like digital art, photo editing, or automated collage generation, where seamless integration of multiple visual elements is essential. The invention improves upon prior methods by introducing flexibility in image positioning, reducing manual adjustments, and enhancing the quality of the final synthetic image.
4. The computer-implemented method of claim 1 , wherein generating a synthetic image for the canvas comprises determining a synthetic to real distance indicative of how close the synthetic image is from a real source image.
This invention relates to computer-generated imagery (CGI) and synthetic image generation, specifically addressing the challenge of creating realistic synthetic images that closely resemble real-world reference images. The method involves generating synthetic images for a digital canvas by determining a synthetic-to-real distance, which quantifies how closely the synthetic image matches a real source image. This distance metric helps assess the realism and fidelity of the generated synthetic image, ensuring it aligns with the desired visual characteristics of the real source. The process may include analyzing visual features, textures, or structural elements to compute this distance, enabling adjustments to improve realism. The method can be applied in various applications, such as digital art, virtual environments, or augmented reality, where realistic synthetic imagery is essential. By dynamically evaluating and refining the synthetic image based on its proximity to the real source, the invention enhances the quality and authenticity of computer-generated visuals.
5. The computer-implemented method of claim 1 , wherein generating a synthetic image for the canvas based on the vector for the canvas comprises generating, in an embedded vector space, a single feature vector based on multiple positional feature vectors for each of the images in the merged space.
This invention relates to generating synthetic images in a computer-implemented system, particularly for applications in image synthesis, machine learning, or digital content creation. The problem addressed is the efficient and coherent generation of synthetic images from multiple input images, ensuring consistency and relevance to a target canvas or output. The method involves creating a synthetic image for a canvas by processing input images in an embedded vector space. First, multiple positional feature vectors are extracted for each input image, representing different aspects or regions of the images. These feature vectors are then combined into a single feature vector in the embedded vector space, which captures the collective characteristics of the input images. This single feature vector is used to generate the synthetic image, ensuring that the output is a coherent representation derived from the merged input data. The approach leverages vector space embedding to maintain semantic and positional relationships between the input images, enabling the generation of high-quality synthetic images that are contextually relevant to the canvas. This method is particularly useful in applications requiring dynamic image synthesis, such as virtual environments, augmented reality, or automated content generation.
6. The computer-implemented method of claim 1 , wherein generating a synthetic image for the canvas comprises obtaining a loss factor from the discriminative tool with the vector for the canvas in the merged space, and modifying the vector for the canvas in the merged space when the loss factor is greater than a selected threshold.
This invention relates to computer-implemented methods for generating synthetic images, particularly in the context of machine learning and generative models. The problem addressed involves improving the quality and accuracy of synthetic image generation by refining the input data before synthesis. The method involves generating a synthetic image for a canvas by first obtaining a vector representation of the canvas in a merged feature space. This vector is then processed using a discriminative tool, which evaluates the vector and outputs a loss factor. If the loss factor exceeds a predefined threshold, the vector is modified to reduce discrepancies or errors. The modified vector is then used to generate the synthetic image, ensuring higher fidelity and consistency with desired characteristics. The discriminative tool acts as an evaluator, assessing the quality or suitability of the vector representation. The modification step adjusts the vector to improve its alignment with the target image properties, enhancing the final output. This approach helps mitigate issues like artifacts, distortions, or mismatches in the generated synthetic images, making the process more robust and reliable. The method is particularly useful in applications requiring high-quality synthetic image generation, such as data augmentation, virtual environments, or creative design tools.
7. The computer-implemented method of claim 1 , wherein generating a synthetic image for the canvas comprises subtracting a spatial feature vector from the vector for the canvas in the merged space to obtain a residual vector, and determining a loss factor using a deep vision tool based on a context of the residual vector.
This invention relates to computer-implemented methods for generating synthetic images, particularly in the context of digital art or design workflows. The problem addressed is the need to efficiently produce high-quality synthetic images that maintain visual coherence with a given canvas while allowing for controlled modifications. The method involves generating synthetic images by manipulating feature vectors in a merged vector space. The process begins by obtaining a vector representation of the canvas in a merged space, which combines multiple feature dimensions. To generate a synthetic image, a spatial feature vector is subtracted from the canvas vector to produce a residual vector. This residual vector captures the differences needed to modify the canvas while preserving its underlying structure. A deep vision tool, such as a neural network, then analyzes the context of the residual vector to determine a loss factor. This loss factor quantifies the deviation introduced by the modification, ensuring that the synthetic image remains visually consistent with the original canvas. The method allows for iterative refinement, enabling users to adjust the synthetic image while maintaining control over the visual outcome. The approach leverages deep learning to automate the generation process while preserving artistic intent.
8. The computer-implemented method of claim 1 , further comprising associating the synthetic image with a synthetic label and storing the synthetic image and the synthetic label in an image database for training the discriminative tool.
This invention relates to generating synthetic images for training discriminative tools, such as machine learning models, in computer vision applications. The problem addressed is the need for large, diverse datasets to train accurate models, which can be costly or difficult to obtain in real-world scenarios. The solution involves creating synthetic images that mimic real-world data, ensuring they are labeled with synthetic labels that correspond to the generated content. The method includes generating a synthetic image using a generative model, such as a generative adversarial network (GAN) or a variational autoencoder (VAE). The synthetic image is then associated with a synthetic label, which may include metadata or annotations describing the image's content, such as object classes, attributes, or spatial relationships. The synthetic image and its corresponding label are stored in an image database, which is used to train a discriminative tool, such as a classifier or detector. This approach allows for scalable and controlled generation of training data, improving model performance without relying solely on real-world datasets. The invention ensures that the synthetic images and labels are structured in a way that is compatible with existing training pipelines, enabling seamless integration into machine learning workflows. By leveraging synthetic data, the method reduces dependency on manual annotation efforts and provides a flexible way to augment training datasets with diverse examples.
9. The computer-implemented method of claim 1 , further comprising modifying a coefficient of the generative tool when the discriminative tool recognizes the synthetic image as synthetic.
A computer-implemented method improves the generation of synthetic images by refining a generative model based on feedback from a discriminative model. The method operates in the domain of generative adversarial networks (GANs), where a generative tool creates synthetic images, and a discriminative tool evaluates their realism. The problem addressed is the persistent generation of unrealistic synthetic images due to suboptimal training of the generative model. The method enhances the generative tool by dynamically adjusting its coefficients when the discriminative tool identifies a synthetic image as synthetic. This adjustment improves the generative model's ability to produce more realistic images over time. The discriminative tool is trained to distinguish between real and synthetic images, providing feedback to the generative tool. The generative tool generates synthetic images by applying learned coefficients to input data, such as noise vectors or latent variables. The method ensures continuous refinement of the generative model by iteratively modifying its coefficients based on the discriminative tool's feedback, leading to progressively more realistic synthetic images. This approach enhances the performance of GANs by improving the generative model's ability to deceive the discriminative model, resulting in higher-quality synthetic outputs.
10. The computer-implemented method of claim 1 , further comprising modifying a coefficient of the discriminative tool to recognize the synthetic image as synthetic.
The invention relates to computer vision and synthetic image detection, addressing the challenge of distinguishing synthetic images from real ones. Synthetic images, often generated by artificial intelligence, can be used for malicious purposes such as deepfakes or misinformation. The method involves training a discriminative tool, such as a machine learning model, to differentiate between real and synthetic images. The discriminative tool is trained using a dataset containing both real and synthetic images, allowing it to learn distinguishing features. Once trained, the tool can analyze new images to determine their authenticity. The method further includes modifying a coefficient of the discriminative tool to improve its ability to recognize synthetic images as synthetic. This adjustment enhances the tool's accuracy in identifying synthetic content, reducing false positives and negatives. The technique is applicable in security, media verification, and content moderation, ensuring reliable detection of AI-generated images. The invention improves upon existing methods by dynamically adjusting the model's parameters to better handle evolving synthetic image generation techniques.
11. A system comprising: one or more processors; and a memory coupled to the one or more processors, the memory including instructions that, when executed by the one or more processors, cause the one or more processors to: receive a compositional constraint from a user for creating a canvas merging multiple images; find a vector for the canvas in a merged space associated with the compositional constraint; generate a synthetic image for the canvas based on the vector for the canvas and a generative tool trained in a generative adversarial configuration against a discriminative tool; and provide the synthetic image for the canvas to a user when the discriminative tool identifies the synthetic image as a real image, wherein to generate a synthetic image for the canvas the one or more processors are configured to merge a first saliency measure for a first image in the canvas with a second saliency measure for a second image in the canvas to form a compound saliency measure associated with the vector for the canvas in the merged space.
This system addresses the challenge of merging multiple images into a cohesive synthetic canvas while adhering to user-defined compositional constraints. The system uses machine learning, specifically a generative adversarial network (GAN), to create a synthetic image that blends multiple input images based on their saliency measures. A user provides compositional constraints, such as desired layout or stylistic preferences, which guide the merging process. The system first computes a vector representing the canvas in a merged space that aligns with these constraints. It then generates a synthetic image by combining saliency measures from the input images, forming a compound saliency measure that informs the generative model. The generative tool, trained in a GAN configuration, produces the synthetic image, which is validated by a discriminative tool. Only when the discriminative tool identifies the synthetic image as real is it provided to the user. This approach ensures the output image maintains visual coherence and adheres to the specified constraints, solving the problem of seamlessly integrating multiple images into a single, aesthetically pleasing composition.
12. The system of claim 11 , wherein to receive a compositional constraint the one or more processors are configured to receive a positional constraint within the canvas for at least one of the images.
This invention relates to digital image composition systems, specifically addressing the challenge of automatically arranging multiple images within a defined canvas while adhering to user-specified constraints. The system processes a set of input images and a target canvas area, generating a composition that optimizes visual appeal and coherence. A key feature is the ability to incorporate compositional constraints, including positional constraints that dictate the placement of specific images within the canvas. These constraints ensure that certain images appear in predefined locations, such as aligning a primary subject to a focal point or maintaining spatial relationships between elements. The system may also apply additional constraints like aspect ratio preservation, overlap minimization, or aesthetic rules (e.g., rule of thirds) to refine the arrangement. The output is a composed image that integrates the input images according to the specified constraints while maintaining visual harmony. This approach is particularly useful in applications like photo collages, automated design tools, or content generation where precise image placement is critical. The system dynamically adjusts the composition to satisfy the constraints while optimizing for overall visual quality.
13. The system of claim 11 , wherein to generate a synthetic image for the canvas the one or more processors are configured to relax a positional condition for at least two images within the canvas.
This invention relates to image synthesis systems for generating synthetic images from multiple input images. The problem addressed is the challenge of combining multiple images into a cohesive synthetic image while maintaining visual consistency and avoiding artifacts. The system uses a canvas as a workspace where input images are placed, and a relaxation process is applied to adjust the positional constraints of at least two images within the canvas. This relaxation allows the system to dynamically modify the placement or alignment of images to improve the overall composition, ensuring smoother transitions and better visual coherence. The system may also include a user interface for selecting and manipulating the input images, as well as a rendering module to produce the final synthetic image. The relaxation process helps avoid rigid positioning conflicts, enabling more natural and aesthetically pleasing results. The invention is particularly useful in applications like image stitching, panoramic generation, or virtual scene construction, where seamless integration of multiple images is critical. The system dynamically adjusts image positions to minimize visual discrepancies, improving the quality of the synthesized output.
14. The system of claim 11 , wherein to generate a synthetic image for the canvas the one or more processors are configured to determine a synthetic to real distance indicative of how close the synthetic image is from a real source image.
This invention relates to image synthesis systems, specifically for generating synthetic images that closely resemble real-world source images. The system addresses the challenge of creating realistic synthetic images by evaluating and optimizing their similarity to real images. The core functionality involves determining a synthetic-to-real distance metric, which quantifies how closely the synthetic image matches the real source image. This metric helps assess the realism of the generated image, ensuring it aligns with the desired visual characteristics of the real source. The system processes input data, such as user-provided parameters or environmental factors, to generate the synthetic image while continuously refining it based on the synthetic-to-real distance. This ensures the output image maintains high fidelity to the real source, improving applications like virtual reality, augmented reality, and digital content creation. The system may also include additional features like real-time adjustments and user feedback integration to further enhance the realism of the synthetic images. By dynamically adjusting the synthesis process based on the synthetic-to-real distance, the system ensures the generated images are both visually accurate and contextually appropriate.
15. The system of claim 11 , wherein to generate a synthetic image for the canvas based on the vector for the canvas the one or more processors are configured to generate, in an embedded vector space, a single feature vector based on multiple positional feature vectors for each of the images in the merged space.
This invention relates to generating synthetic images for a canvas using vector-based image processing. The system addresses the challenge of creating visually coherent synthetic images by leveraging embedded vector spaces to combine multiple images into a unified representation. The system first merges multiple images into a shared vector space, where each image is represented by a positional feature vector. These positional feature vectors are then aggregated into a single feature vector in the embedded vector space. This single feature vector is used to generate a synthetic image for the canvas, ensuring consistency and coherence across the merged images. The system may also include a user interface for selecting and arranging images, as well as a display for presenting the synthetic image. The invention improves upon traditional image synthesis methods by using vector-based representations to enhance visual continuity and reduce artifacts in the generated output. The approach is particularly useful in applications requiring seamless blending of multiple images, such as digital art, virtual environments, or augmented reality.
16. The system of claim 11 , wherein to generate a synthetic image for the canvas the one or more processors are configured to obtain a loss factor from the discriminative tool with the vector for the canvas in the merged space, and to modify the vector for the canvas in the merged space when the loss factor is greater than a selected threshold.
This invention relates to a system for generating synthetic images using machine learning techniques, specifically in the domain of generative adversarial networks (GANs) or similar deep learning models. The system addresses the challenge of improving the quality and realism of synthetic images by refining input data through a discriminative tool that evaluates and adjusts image representations in a merged feature space. The system includes one or more processors configured to process input data, such as a vector representing a canvas or image, within a merged space that combines features from multiple sources. A discriminative tool, such as a discriminator network in a GAN, assesses the vector to determine its quality or realism. The tool generates a loss factor, which quantifies deviations from desired characteristics. If the loss factor exceeds a predefined threshold, the system modifies the vector to improve its representation. This adjustment ensures the generated synthetic image meets quality standards before final output. The system may also include additional components, such as a generator network that produces the synthetic image from the refined vector. The discriminative tool may operate iteratively, refining the vector through multiple cycles until the loss factor falls within acceptable limits. This approach enhances the fidelity of synthetic images, making them more suitable for applications like data augmentation, artistic generation, or virtual environment design. The invention focuses on dynamic refinement of image representations to achieve higher-quality outputs.
17. The system of claim 11 , wherein to generate a synthetic image for the canvas the one or more processors are configured to subtract a spatial feature vector from the vector for the canvas in the merged space to obtain a residual vector, and determining a loss factor using a deep vision tool based on a context of the residual vector.
The invention relates to a computer vision system for generating synthetic images, particularly in the context of image editing or augmentation. The system addresses the challenge of accurately modifying or extending an existing image while maintaining visual coherence and realism. The core functionality involves processing an input image (referred to as a "canvas") by transforming it into a high-dimensional vector representation in a merged feature space. This space combines multiple spatial and contextual features extracted from the image. To generate a synthetic image, the system subtracts a spatial feature vector from the canvas's vector representation, producing a residual vector. This residual vector captures the differences or modifications needed for the synthetic image. A deep vision tool, such as a neural network, then analyzes the residual vector's context to determine a loss factor, which quantifies the deviation or error introduced by the modification. This loss factor is used to refine the synthetic image, ensuring it aligns with the original image's visual characteristics. The system may also include preprocessing steps to extract and merge spatial features from the input image, as well as post-processing to apply the synthetic modifications. The overall goal is to enable realistic image synthesis while preserving structural and contextual integrity.
Unknown
June 30, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.